Session-based recommender systems are crucial for generating recommendations based on session data. These systems analyze session data to understand user preferences and tailor recommendations accordingly. Generating diverse recommendations is essential for enhancing user satisfaction by allowing users to explore various items. Additionally, items with fewer interactions often suffer from long-tail issues, where they are rarely recommended. As a result, users do not get the opportunity to discover these items, leading to limited exposure. Furthermore, this can result in a reduced overall interaction with the application. This work aims to generate diverse recommendations while addressing long-tail issues. To tackle these challenges, this work proposes a methodology combining Natural Language Processing and Item-based Collaborative Filtering to create diversified recommendations while mitigating Long-tail problems (NIDL). The proposed method was evaluated using the MovieLens 100 K and 1 M benchmark datasets, demonstrating superior performance compared to current leading models in terms of both accuracy and diversity.